Source: Nature
The pipeline problem is that students using AI tools during their most formative research years are statistically more likely to leave academia entirely. This suggests the technology shifts how people evaluate the value of deep disciplinary work before they've completed it. For fields dependent on continuous generational knowledge transfer, this matters: if early AI adoption correlates with early exit, universities face a shrinking talent pool when they most need mentors for the next cohort of researchers. The cost isn't efficiency or output quality. It's the erosion of institutions whose legitimacy depends on people willing to work through pre-AI modes of knowledge production.